EE Test Engineer Cheat Sheet

The bench: which tool does what

Instrument	What it shows	Typical uses
Oscilloscope + probes	Voltage vs. time	Edges, ripple, transients, step response, eye diagrams	Try it
DMM (multimeter)	Accurate DC/AC volts, amps, ohms	DC levels, regulation, quiescent current, continuity	Try it
Spectrum analyzer	Power vs. frequency	Harmonics, spurs, noise floor, EMI pre-scan	Try it
VNA (vector network analyzer)	S-parameters vs. frequency	Cables, filters, connectors, impedance matching	S11 plot
BERT / pattern generator	Bit errors over millions of bits	BER, jitter tolerance, link margin	BER plot
Programmable power supply	Controlled input voltage	Line regulation, brown-out, power sequencing	Try it
Electronic load	Controlled current draw	Load regulation, load-step transient response	Try it
Power analyzer	True input and output power	Efficiency, power factor, standby power
Current probe (Hall / clamp)	Current vs. time on a scope	Inrush, ripple current, switching current
Function generator / AWG	Stimulus signals	Frequency response, step inputs, test patterns	Try it
TDR (scope or VNA mode)	Impedance vs. distance	Finding discontinuities on traces and cables
Thermal camera / thermocouple	Temperature	Hot spots, derating, thermal validation

Inside a scope probe

The probe is part of the circuit you're measuring. Understand it and half of all "weird scope readings" explain themselves.

• It's a 10:1 divider. 9 MΩ in the tip and the scope's 1 MΩ input. Set the channel to 10× (or let it auto-detect), or every reading is off by 10.
• Compensate it. Clip onto the scope's calibration output and turn the trim until the square wave has flat tops. Redo it whenever the probe moves to another channel or scope.
• It loads the circuit. The tip looks like ~10 MΩ in parallel with ~10 pF. At 100 MHz, 10 pF is only ~160 Ω, so the probe becomes the load. For fast signals use an active probe (~1 pF).
• The ground lead rings. Its inductance plus the tip capacitance form an LC circuit, which adds fake ringing on fast edges. The short ground spring fixes it.
• 1× probes put the whole cable capacitance on the circuit, and their bandwidth is often under 20 MHz. Use them only for small, slow signals.
• 50 Ω input mode is for coax and active probes only. It is usually limited to about 5 Vrms: check the label before connecting.

Digital and high-speed signals

The question here is always: does each bit arrive on time and at a clear level?

Measurement	What it means	How to measure	Tip
Rise / fall time	Time for an edge to go 10% → 90% (or 20–80%). Too slow = timing errors; too fast = ringing and EMI.	Scope auto-measure	Scope + probe bandwidth should be 3–5× faster than the edge. Scope rise time ≈ 0.35 / BW.
Overshoot / ringing	Edge shoots past its final level and oscillates. Usually impedance mismatch or stray inductance.	Scope, as % of amplitude	Use the short ground spring, not the long ground lead (the lead itself rings).
Timing: delay, skew, setup/hold	When edges arrive relative to each other or to the clock.	Scope, 2+ channels, measure 50% to 50%	Deskew probes first; cables add about 5 ns per metre.
Jitter	Edges arriving early or late vs. the ideal time. Random jitter (RJ) is unbounded; deterministic (DJ) is bounded (data-dependent, crosstalk, periodic).	Scope jitter analysis (TIE histogram); phase-noise analyzer for clocks	Total jitter is quoted at a BER: TJ ≈ DJ + 14 × RJ(rms) at 10⁻¹².
ISI (inter-symbol interference)	Earlier bits smear into the current bit because the channel loses high frequencies (trace/cable loss) and reflects. Shows up as data-dependent jitter and a closing eye; worst for a lone bit after a long run of 1s or 0s.	Eye diagram and DDJ at the receiver with a long pattern (PRBS31); channel loss as VNA S21 at the Nyquist frequency (bit rate ÷ 2)	If the PRBS31 eye is much worse than PRBS7, ISI is the culprit. The fix is emphasis / equalization (last row).
Eye diagram	All bits overlaid. Open eye = healthy link. Height = noise margin, width = timing margin.	Scope eye mode with clock recovery, checked against a mask	Any mask hit = fail. Measure at the receiver end of the channel.
BER (bit error rate)	Errors ÷ bits sent. Typical spec: 10⁻¹².	BERT sending a PRBS pattern (PRBS7, PRBS31)	To claim 10⁻¹² at 95% confidence with zero errors you need ~3 × 10¹² bits (about 5 min at 10 Gb/s).
Pre-/de-emphasis, EQ	The transmitter boosts transitions to pre-compensate cable loss at high frequency; the receiver's CTLE/DFE equalizer does the same at the far end.	Sweep emphasis/EQ settings, compare eye opening or BER	More is not better: too much emphasis closes the eye again. Pick the setting with best margin.

Power electronics

The question: is the output clean, stable and efficient across every input voltage and every load?

Measurement	What it means	How to measure	Tip
Ripple and noise	The AC wiggle riding on the DC output, mostly at the switching frequency, plus spikes. Spec'd in mV peak-to-peak.	Scope, AC-coupled, 20 MHz bandwidth limit, probe right at the output capacitor	A long ground lead picks up switching noise and shows fake ripple: use a ground spring or tip-and-barrel. Measure at full load.
Line regulation	How much Vout moves when Vin moves. % = ΔVout ÷ Vout(nominal) × 100, over the full Vin range.	Programmable supply sweeps Vin min → max at fixed load; DMM on Vout	Sense voltage at the board (4-wire / Kelvin) so cable drop doesn't count.
Load regulation	How much Vout moves from no load to full load. % = (V(no load) − V(full load)) ÷ V(nominal) × 100.	Electronic load steps through load points; DMM on Vout	Same 4-wire rule. Let it warm up before reading.
Load transient response	Dip or overshoot and recovery time when the load changes fast (e.g. 10% → 90%).	E-load in dynamic mode (set the slew rate), scope on Vout, trigger on the load current	Report peak deviation (mV) and time to get back inside the band. Ringing hints at a weak control loop.
Efficiency	η = Pout ÷ Pin × 100%. The rest turns into heat.	Power analyzer, or 4 DMMs: Vin, Iin, Vout, Iout	Plot it from 10% to 100% load; light load is often worst. Wait for thermal steady state.
Inrush current	The current spike at turn-on while input capacitors charge. Can trip breakers, blow fuses, weld relay contacts.	Current probe on the input, scope in single-shot mode, trigger on rising current	Fully discharge caps between tries. Worst case: max Vin, or AC switched on at the voltage peak.
Quiescent / standby current	Current drawn with no load. Decides battery life.	DMM in series on the µA range	Watch the meter's burden voltage on low ranges.
Protections and start-up	Does it shut down safely on short circuit, overvoltage, overtemperature? Do rails power up in the right order?	E-load short mode; multi-channel scope on all rails	Classic interview topic: OCP, OVP, OTP, UVLO, soft-start, sequencing.

Analog and RF

The question: how much of what comes out is the real signal, and how much is noise, distortion or loss?

Measurement	What it means	How to measure	Tip
SNR (signal-to-noise ratio)	Signal power ÷ noise power, in dB. Higher = cleaner.	Spectrum analyzer, or FFT of a scope/ADC capture: signal peak vs. total noise in the band	Always state the bandwidth: more bandwidth = more noise. Ideal ADC: SNR = 6.02 × bits + 1.76 dB.
Noise floor	The lowest level the system can see: the "grass" at the bottom of the spectrum.	Spectrum analyzer with the input terminated in 50 Ω	Narrower RBW lowers the displayed floor (10× narrower = −10 dB). Physics limit: −174 dBm/Hz at room temperature.
THD, SINAD, ENOB	THD = harmonics vs. the fundamental. SINAD = signal vs. noise + distortion. ENOB = (SINAD − 1.76) ÷ 6.02.	Very clean sine in (low-distortion generator + filter), FFT the output	Use a window (Blackman-Harris) or coherent sampling, or the FFT smears and lies.
Bandwidth / frequency response	Gain vs. frequency. Bandwidth = where gain drops 3 dB.	Swept sine (generator + scope), Bode analyzer, or VNA S21	Sweep log-spaced points; check phase too, not only gain.
Spectrum, spurs, EMI	Which frequencies are present: clocks, switching harmonics, interference.	Spectrum analyzer; near-field probes to find the source	If a spur changes when you add input attenuation, the analyzer is overloaded, not the DUT.
S-parameters	How a device reflects and passes RF. S11 = input match, S21 = gain/loss through, S22 = output match, S12 = reverse isolation.	VNA, calibrated (SOLT: short, open, load, thru) at the cable ends	Recalibrate whenever cables, adapters or the frequency span change. S11 of −20 dB ≈ VSWR 1.22.

Control systems

The question: when the setpoint or load changes, does the loop get there fast, without overshooting too much, and stay there? Almost everything comes from one test: apply a step and record the output.

Measurement	What it means	How to measure	Tip
Overshoot	How far the output goes past its final value. % = (peak − final) ÷ final × 100.	Step the setpoint (or the load); capture on scope or data logger; read the peak	More overshoot = less damping = closer to instability.
Undershoot	The output dips below the final value after the overshoot (or first moves the wrong way).	Same capture	Several big swings back and forth = ringing = a lightly damped loop.
Rise time	10% → 90% of the final value: how fast the loop reacts.	Scope cursors on the step capture	Faster usually costs overshoot: it's a trade-off.
Settling time	Time until the output enters the band (±2% or ±5%) and never leaves again.	Same capture, measured to the last time it exits the band	Rough estimate: 4 ÷ (ζωn) for the 2% band.
Steady-state error	Gap between setpoint and where the output finally rests.	Averaged DMM or logger reading after settling	Integral action (the I in PID) drives it to zero.
Gain and phase margin	How far the loop is from oscillating. Common targets: phase margin ≥ 45°, gain margin ≥ 6–10 dB.	Frequency-response (Bode) analyzer injecting a small signal across a 5–50 Ω resistor in the feedback path	This is how power-supply loops are checked. ~45° PM gives roughly 20–25% overshoot; ~60° about 10%.

Bench tips that interviewers love

A good test engineer always asks: could my setup be causing what I see?

• Shortest ground possible. The long probe ground lead is an antenna and an inductor: it adds ringing and fake noise. Use the ground spring.
• Compensate 10× probes on the scope's calibration square wave before measuring.
• Enough bandwidth, but no more than needed. Scope at 3–5× the signal's fastest content; turn on the 20 MHz limit for ripple.
• Probes load the circuit. A 10× passive probe is about 10 MΩ in parallel with ~10 pF. For fast or high-impedance nodes use an active probe.
• Never float the scope (cutting the ground pin) to measure a switch node or high side. Use a differential probe.
• 4-wire (Kelvin) sensing for low resistance, regulation and efficiency: cable drop otherwise ends up in your result.
• Averaging removes random noise only. Don't average when you're measuring noise or jitter.
• Warm up instruments (~30 min), check calibration dates, and let the DUT reach thermal steady state.
• Guard-band your limits. Test limit = spec limit minus measurement uncertainty, so a pass is really a pass.
• Write the setup down: instruments, settings, temperature, photo of the wiring. Repeatable or it didn't happen.
• Safety: discharge capacitors, use an isolation transformer for mains work, wear an ESD strap.
• Automate. Most instruments speak SCPI over USB/LAN/GPIB; Python + PyVISA scripts are a big plus in test roles.

Quick formulas

Formula	When you need it
dB (power) = 10 log(P2 ÷ P1); dB (voltage) = 20 log(V2 ÷ V1)	Everything in dB. 3 dB = 2× power, 6 dB = 2× voltage, 20 dB = 10× voltage.
0 dBm = 1 mW ≈ 0.224 Vrms into 50 Ω; dBµV = dBm + 107 (50 Ω)	Reading a spectrum analyzer or EMI receiver
Vrms = Vpeak ÷ √2 (sine only)	DMM AC readings
Scope rise time ≈ 0.35 ÷ bandwidth	Is my scope fast enough?
Nyquist: f_max = sample rate ÷ 2	Will my scope or ADC alias this signal?
SNR(ideal ADC) = 6.02 × N + 1.76 dB	ADC testing, ENOB
Thermal noise = √(4kTRB); −174 dBm/Hz at 290 K	The lowest noise floor possible
Γ = (ZL − Z0) ÷ (ZL + Z0); VSWR = (1 + |Γ|) ÷ (1 − |Γ|)	Reflections, impedance matching
Bits to test ≈ 3 ÷ target BER	BER at 95% confidence with zero errors
η = Pout ÷ Pin; Ploss = Pin − Pout	Efficiency and heat
Meter error ≈ Rsource ÷ (Rsource + Rin)	Loading by a DMM or probe
Uncertainty (independent errors) = √(e1² + e2² + …)	Uncertainty budgets

Adapted from the "EE Test Engineer Cheat Sheet" (Seba, 3 Nov 2026), with links into this tool.